STEPS Survey monitoring report
06 December, 2024
The fieldwork for the STEPS survey in
Country Name started on 2018-06-09. Up to date, 2553 records for STEP 1
and 2499 records for STEP3 completed interviews have been exported to
the central server. Sample size = 3000.
Number of records
Cumulative nr of records
Map
Records by Age and Sex
Sample coverage
Records by cluster
The table below shows the number of records from Step 1 per Cluster. According to sampling strategy for Country Name STEPS survey there should be 11 subjects selected in each Cluster. All Clusters in the table with more than 11 records will be highlighted in red.
Records by Device ID
Number of records by Device ID will increase in time. The following Table shows the cumulative nuber of records for Step 1 and Step 2. The number of entries can be changed from the drop down menu and number of records can be sorted by pressing on the sort button on the top.
Records per day by device
In the chart below shows the number of interviews per day for all devices (data collectors). Bubbles will increase in size and change color proportional to number of interviews per day. Bubbles colored in yellow, orange or red will represent cases when data collectors conducted more than 4 interviews, that is usually not possible to achieve during one day.
QR codes
QR codes are used in STEPS survey as unique identifiers and have to be scanned at each step. Data collectors are not supposed to scan the same QR code for two respondents. In case if it happens the records must be corrected immediately on the server side by editing the wrong records. The tables below will show all duplicates for QR codes at the moment for Step 1 and Step 3 separately, if there are any.
Duplicate QR Codes STEP1
| QR | Device.IDs | Count |
|---|---|---|
| 86492502 | 114, 114 | 2 |
| 54228775 | 114, 114 | 2 |
| Total | 4 |
Duplicate QR Codes STEP3
| QR | Device.IDs | Count |
|---|---|---|
| Total | 0 |
QR STEP3 ≠ QR STEP1
As mentioned above, the QR codes serve as unique identifiers and are used to match data collected at different steps. Step 3 data is only collected on respondents who participate in Step 1 and the QR codes from Step3 shall match the codes from Step 1. In case if they don’t, the list of unmatched codes will appear in the table below.| QR3 | Device.ID | Count |
|---|---|---|
| 35197807 | 114 | 1 |
| 59893105 | 114 | 1 |
| 97415581 | 114 | 1 |
| Total | 3 |
Average HH size
Household listing and selection in STEPS survey is done during the
field work and data collectors have to enumerate all eligible household
members and then randomly select one. If ennumerators don’t list all
eligible members that may introduce bias to selection process and have
an impact on the survey results. The bar plot below shows the average
number of eligible household members by Device ID. When data collection
is conducted in multiple households, the average shall be close to the
national estimated average. The green
line represents the expected country average that was estimated to be ~
2.6 members per household. The estimated average can be modified in
future reports if needed.
Timing
The duration of the interview is critical for the quality of data collected in the STEPS survey. Interviewers must allocate sufficient time to ensure respondents fully understand the questions and provide accurate answers. The expected interview time can vary depending on the inclusion of optional modules or country-specific questions. If an interview is found to be shorter than the expected duration, those instances must be reviewed and addressed accordingly.
The chart below displays the average interview times for Step 1 and Step 3 per week for the entire survey. We do not expect to see significant weekly variations, particularly a decrease in the average interview time over time.
Interview time by week
The next chart shows the average interview times per Device ID/interviewer. The orange lines represent the average interview time for all interviewers.
Average interview time
Blood pressure measurement is a crucial part of Step 2 data
collection and must be conducted according to the recommended procedure.
Data collectors are required to measure each participant’s blood
pressure three times, with at least a 3-minute interval between
measurements.
The graph below illustrates the average BP measurement waiting time for each interviewer (Device ID). The size of the bubbles corresponds to the number of interviews conducted. If the average waiting time is less than 3 minutes, the bubbles will appear in red.
BP average waiting time
Frequencies
The recorded dates and months of birth should generally be evenly distributed when computing average numbers. If frequencies for certain dates or months are much higher, this needs to be investigated and addressed. However, it is possible that some respondents may not recall their exact date or month of birth, which could result in higher frequencies for specific dates or months.
Birth Date and Month Frequencies
Frequency of physical measurement results
The STEPS survey, a standardized method for collecting, analyzing, and disseminating data on noncommunicable diseases (NCDs) and their risk factors, benefits significantly from incorporating objective physical measurements. These measurements provide a robust foundation for understanding the prevalence and distribution of NCDs, thereby informing public health policies and interventions. They also enable the detection of subtle trends over time that might be missed with self-reported data alone. It is crucial, however, that physical measurements are conducted in a standardized manner according to the predefined methodology.
The purpose of the graphs below is to highlight any potential issues related to the measurements conducted during Step 2 by data collectors:
- Identification of Outliers: Unusually high or low readings can indicate potential data entry errors or anomalies that need further investigation.
- Visualizing Frequency Distribution: the graphs can detect any unusual patterns or irregularities. For instance, a high frequency of specific values might suggest rounding errors, measurement bias, or the recording of reported values instead of measured ones.
- Detecting Measurement Clusters: Clusters of measurements at certain value levels can indicate that some interviewers might not be following the proper measurement protocols. This clustering needs to be investigated and corrected to ensure accurate data collection.
Last Digits Frequency
The last digits frequency analysis for physical measurements, such as blood pressure height, weight, and other readings, is conducted to evaluate the distribution of the final digit in recorded values. This type of analysis helps identify potential data quality issues, such as rounding errors or systematic biases in the measurement process.
Detection of Rounding Bias:
Uniform Distribution Expectation: In a perfectly unbiased and accurate measurement process, the last digits (0-9) should be uniformly distributed. Each digit should appear with approximately equal frequency. Rounding Patterns: If certain digits (e.g., 0 or 5) appear more frequently than others, it may indicate that data collectors are rounding values to these digits rather than recording the exact measurements. This can compromise the accuracy of the data. Quality Control:
The bar plots below will help to:
Identify anomalies: By examining the last digits, researchers can detect anomalies that suggest deviations from standard measurement protocols. For example, a disproportionate frequency of specific digits may indicate that some data collectors are not following proper procedures or are recording of reported values instead of measured ones.
Highlight the need for additional training or calibration of devices: Frequent irregularities in last digit distribution can highlight the need for additional training for data collectors or recalibration of measurement devices to ensure accurate readings.
Ensure reliable and precise measurements: Crucial for the integrity of the data analysis. Ensuring that the last digits are evenly distributed helps maintain the validity of statistical analyses and research conclusions.
Trust in Findings: Stakeholders, including policymakers and healthcare providers, rely on accurate data to make informed decisions. Demonstrating that data collection processes are robust and free from bias enhances the credibility of the findings.
